e. levin, j. miller and a. prygarin-summing pomeron loops in the dipole approach
TRANSCRIPT
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
1/36
arXiv:0706
.2944v3
[hep-ph]
23Oct2007
Preprint typeset in JHEP style - HYPER VERSION TAUP -2858-07
February 1, 2008
Summing Pomeron loops in the dipole approach
E. Levin J. Miller and A. Prygarin
Department of Particle Physics, School of Physics and Astronomy
Raymond and Beverly Sackler Faculty of Exact Science
Tel Aviv University, Tel Aviv, 69978, IsraelDepartment of Particle Physics, University of Santiago de Compostela,
Santiago de Compostela, 15782, Spain
Abstract: In this paper we argue that in the kinematic range given by 1 ln(1/2S) S Y 1S , we canreduce the Pomeron calculus to the exchange of non-interacting Pomerons with the renormalized amplitude of their
interaction with the target. Therefore, the summation of the Pomeron loops can be performed using the improved
Mueller, Patel, Salam and Iancu approximation and this leads to the geometrical scaling solution. This solution is
found for the simplified BFKL kernel. We reproduce the findings of Hatta and Mueller that there are overlapping
singularities. We suggest a way of dealing with these singularities.
Keywords: BFKL Pomeron, Pomeron loops, Mean field approach, Exact solution.
[email protected], [email protected];[email protected];[email protected] address
http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://arxiv.org/abs/0706.2944v3http://jhep.sissa.it/stdsearchhttp://jhep.sissa.it/stdsearchhttp://arxiv.org/abs/0706.2944v3 -
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
2/36
Contents
1. Introduction 1
2. The BFKL calculus in zero transverse dimension: non-interacting Pomerons and improved
Mueller-Patel-Salam-Iancu approach 3
3. Calculation of the simplest diagrams in the BFKL Pomeron calculus 6
3.1 The functional integral formulation of the BFKL Pomeron calculus 7
3.2 Triple Pomeron vertex in momentum representation 8
3.3 The simplest fan diagram 11
3.3.1 The single Pomeron exchange 11
3.3.2 The first fan diagram 12
3.4 The first enhanced diagram 16
3.5 Reduction of the emhanced diagrams to the system of non-interacting Pomerons 18
4. General solution for the simplified BFKL kernel 20
4.1 The main idea 20
4.2 The simplified BFKL kernel and MFA solution 21
4.3 Improved MPSI solution for Pomeron loops 23
4.4 Self-consistency check 25
5. Conclusions 26
1. Introduction
In this paper we discuss a solution for the high energy scattering amplitude in the case of high density QCD. We
would like to go beyond the mean field approximation, where the solution has been discussed and well understood
both analytically and numerically (see [1, 2, 3, 4, 5, 6, 7, 8]), and to approach the summation of the so called Pomeron
loops that should be taken into account [9, 10, 11, 12, 13]. The problem of taking into account the Pomeron loops
can be reduced to the BFKL Pomeron calculus [14, 15, 16] and/or reduced to the solution of the statistical physicsproblem, i.e. the Langevin equation and directed percolation [17, 18, 19].
At the moment the problem of the high energy amplitude in QCD is well understood and can be solved only in
the kinematic region given by
1 ln(1/2S) S Y 1
S(1.1)
1
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
3/36
To go beyond this region we need to know the corrections of the order 2S to the BFKL kernel as well as the corrections
to the vertices of the Pomeron interaction. Eq. (1.1) will play a significant role in our discussion and we wish to
present a more detailed derivation. It is well known that the exchange of one BFKL Pomeron leads to a contribution
which can be written as
A (one Pomeron exchange)
2
S
eY (1.2)
where Y = ln(s/s0) ( s = W2 and W is the energy of the scattering in the c.m. frame ). The intercept of the BFKL
Pomeron can be represented in the form
= S LO BFKL ( = 1/2) + 2S NLO BFKL ( = 1/2) (1.3)
where is the Mellin transform of the BFKL kernel (see more details in [20] ).
The contribution from the exchange of two BFKL Pomerons is proportional to
A (two Pomeron exchange) 2S eY2 (1.4)Comparing Eq. (1.2) to Eq. (1.4) one can see three different kinematic regions:
1. 2 ln(1/S) > Y 1 ;In this kinematic region we need to take into account the BFKL Pomeron at leading order since 2S Y 1and neglect the contribution from multi Pomeron exchanges.
2. 1/S Y 2 ln(1/S) ;We need to take into account multi Pomeron exchange but we can still restrict ourselves to the BFKL kernel
in the leading order approximation.
3. Y 1/S;In this region we have to calculate the next-to-leading BFKL corrections to the BFKL kernel as well as the
corrections to the Pomeron vertices. In addition, it is not clear whether or not we can rely on the BFKLPomeron calculus in this region [21].
Concentrating on the kinematic region of Eq. (1.1) we develop a strategy which consists of three steps. First,
we show in this paper that the BFKL Pomeron interaction in this kinematic region can be reduced to the exchange
of non-interacting Pomerons if we renormalize the low energy amplitude of the interaction of wee partons with the
target.
Using this observation we propose an improved Mueller-Patel-Salam-Iancu method (MPSI) for summation of
Pomeron loop diagrams. This method is actually the t-channel unitarity constraints re-adjusted in a convenient form
for use in the dipole approach of QCD [22, 23].
Finally we propose that the answer which we obtain, is the real solution to our problem at ultra high energy.
We will justify and show that this is a reasonable hypothesis.This paper is organised in the following way. In the next section we discuss the simple case of the BFKL Pomeron
calculus in zero transverse dimensions. Being the simplest model for the Pomeron interaction this approach allows
us to discuss our main ideas and suggestions without complex calculations. Actually, the main content of this section
has been discussed in [24], but for completeness we present this model in our paper. We hope that the reader will
find this instructive in later sections.
2
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
4/36
0
Y
R
Y
R = + 2
Figure 1: The renormalization procedure in the case of the simplest fan diagram.
In section 3 we argue that the main properties of the BFKL Pomeron calculus in zero transverse dimension are
inherent for the BFKL Pomeron calculus in QCD. In particular, we can consider the scattering amplitude in the
kinematic region of Eq. (1.1) as the exchange of BFKL Pomerons neglecting their mutual interactions.
Section 4 is devoted to the calculation of the scattering amplitude in the model for the BFKL kernel which has
been developed in Ref. [25]. In this section we demonstrate how the MPSI approach works and we obtain a formula
for the scattering amplitude. We advocate that the resulting scattering amplitude satisfies the unitarity constraints
both in the s and t channels and could be a good candidate for the answer outside the kinematic region given byEq. (1.1).
In our conclusion we summaries our results and compare them with the approaches that we currently have on
the market.
2. The BFKL calculus in zero transverse dimension: non-interacting Pomerons and
improved Mueller-Patel-Salam-Iancu approach
In this section we analyse the Pomeron diagrams in the BFKL Pomeron calculus in zero transverse dimensions (toy
model of [22, 26, 27]). Recently, this model has become rather popular (see [28] and [29] where the most important
aspects of this model have been discussed and solved)Firstly we consider the simplest fan diagram of Fig. 1 and Fig. 2. It can be calculated in an obvious way, using
the explicit expression for the Pomeron Green function, namely, G(Y Y) = exp( (Y Y)). Indeed, for thediagrams of Fig. 1 we have
A (Fig. 1) = G(Y 0) 2Y0
d Y G(Y Y) G2(Y 0) (2.1)
= eY 2Y0
d Y e(Y+Y) (2.2)
= eY 2
1
e2Y 1
eY
= 2
e2Y
+ ( + 2
) eY
= 2
e2Y
+ R eY
where is the Pomeron intercept which is equal to the triple Pomeron vertex (1 2) in this oversimplified model,and is the amplitude of the Pomeron interaction with the target.
For the diagrams of the second order, given by Fig. 2, we have to integrate over the two rapidity variables y1and y2. The result is
3
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
5/36
Y
0
Y2
Y1
(3)R
(3)R
= + 2 + 3(2) = + 2
(2)R
(2)R
Figure 2: The renormalization procedure in the case of the fan diagram of the second order.
A (Fig. 2) =
= 2 2 3Y0
d y1
y10
d y2 G(Y y1) G(y1 0) G(y2 0) G2(y2 0) (2.3)
= 2 2 3Y0
d y1
y10
d y2 e(Y+y1+y2) = 2 2 3
1
2 2e3Y 1
2e2Y +
1
2 2eY
As one can see, the integration over Y in Fig. 1 and over y1 and y2 in Fig. 2 reduces these two diagrams tothe sum over diagrams which describes two contributions: the exchange of two non-interacting Pomerons and the
exchange of one Pomeron with the renormalized vertices: (2)R = +
2 and (3)R = +
2 + 3
Adding the two contributions of these diagrams gives
A (Fig. 1) + A (Fig. 2) = (2.4)
= 3 e3Y (2 + 2 3) e2Y + ( + 2 + 3) e3Y = 3 e3Y ((2)R )2 e2Y + R eY
Therefore, one can see that the scattering amplitude can be rewritten as the exchange of Pomerons without interaction
between them but with the renormalized Pomeron - particle vertex. In the dipole model this vertex is the amplitude
of the two dipole interaction in the Born approximation of perturbative QCD.
These two examples illustrate our main idea: the BFKL Pomeron calculus in zero transverse dimensions can be
viewed as the theory of free, non-interacting Pomerons whose interaction with the target has to be renormalized.
The master equation for the scattering amplitude in the mean field approximation can be easily rewritten (see Ref.
[24]) in the form N0 (R|Y)
Y= (1 2) R N0 (R|Y)
R(2.5)
with
R =
1 (2.6)
Thus we have shown in Eq. (2.1) and Eq. (2.3) how Eq. (2.6) has started to build up in the perturbative expansion.
The general solution of Eq. (2.5) is the system of non - interacting Pomerons and the scattering amplitude can
be found in the form
N0 (R|Y) = n=1
(1)n Cn nR Gn(Y 0) (2.7)
where the coefficients Cn could be found from the initial conditions, namely, from the expression for the low energy
amplitude. In particular, the initial condition
N0 (R|Y = 0) = = R/(1 + R) (2.8)
4
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
6/36
generates Cn = 1 and the solution is
N0 (R|Y) = R eY
1 + R eY(2.9)
The initial condition of Eq. (2.8) has very simple physics behind it that has been discussed in Ref.[24].
For the analysis of the enhanced diagrams we start from the first diagram of Fig. 3, which can be written asfollows
A (Fig. 3) = (2.10)
2 2Y0
d y1
y10
d y2 G(Y y1) G2(y1 y2) G(y2 0)
= 2 2Y0
d y1
y10
d y2 e(Y+y1y2)
= 2 e2Y + 2 eY + 2 Y eY
where (2 1 ) = 2 (see Ref.[11]).Adding Eq. (2.10) to the exchange of one Pomeron we obtain the following
One Pomeron exchange + A (Fig. 3) = R eR Y 2 e2Y (2.11)
with
R = (2) = + 2 ; R = + ; (2.12)
Therefore, the Pomeron loops can be either large ( where their size in rapidity is of the order of Y) and they can be
considered as un-enhanced diagrams; or small (where their sizes are of the order 1 /) and they can be treated as
the renormalisation of the Pomeron intercept.
In QCD S while 2S . Therefore, the renormalisation of the Pomeron intercept is proportionalto 3S . We can neglect this contribution since (i) there a lot of
2S corrections to the kernel of the BFKL equation
that are much larger than this contribution; and (ii) in the region of Y 1/2S , where we can trust our Pomeron
calculus (see introduction) (R ) Y 1.Concluding this analysis we can claim that the BFKL Pomeron calculus in zero dimension is a theory of non-
interacting Pomerons with renormalised vertices of the Pomeron-particle interaction. In the dipole language, it means
that we have a system non-interacting Pomerons with a specific hypothesis on the amplitude of the dipole interactions
at low energy. For the problem that we are solving here, namely, when we have one bare Pomeron at low energy,
this amplitude is determined by Eq. (2.8).
For such a system we can calculate the scattering amplitude using a method suggested by Mueller, Patel, Salam
and Iancu and developed in a number of papers (see Refs.[23, 36, 32, 37, 35, 38] and references therein). In this method
the scattering amplitude is calculated using the t-channel unitarity constraints which is written in the following way
(assuming that the amplitudes at high energy are purely imaginary, N = Im A):
N([. . . ]
|Y) = N([. . . ]
|Y
Y; P
nP) N([. . . ]|Y
; P
nP) (2.13)
where
stands for all necessary integrations while [. . . ] describes all quantum numbers (dipole sizes and so on ).
The correct implementation of Eq. (2.13) leads for our case to the following formula (see also Refs. [ 35, 36, 30])
NMPSI0 (Y) = 1 exp
BA
(1)R
(2)R
NMFA
(
(1)R |Y Y
NMFA
(
(2)R |Y
|
(1)R
= (2)R
=0(2.14)
5
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
7/36
where NMFA (Y, R) is given by Eq. (2.9)(see also Eq. (2.5)) in the mean field approximation and BA 2S is
the scattering amplitude at low energies which is described by the Born approximation in perturbative QCD. The
difference between Eq. (2.14) and the original MPSI approach is the fact that this equation does not depend on the
value of Y and, because of this, we do not need to choose Y = Y/2 for the best accuracy.
Substituting Eq. (2.9) in Eq. (2.14) we obtain
NMPSI0
BA|Y = 1 exp 1BAe(12)Y
1
BAe(12)Y
0,
1
BAe(12)Y
(2.15)
(0, x) is the incomplete gamma function (see formulae 8.350 - 8.359 in Ref. [40]).
We claim that Eq. (2.15) is the solution to our problem. One can easily see that N0 (|Y) 1 at high energiesin contrast to the exact solution (see [30]). The exact solution leads to the amplitude that vanishes at high energy
(see Refs.[29, 39]). As has been mentioned the solution depends crucially on the initial condition for the scattering
amplitude at low energies. For Eq. (2.15) this amplitude is equal to
NMPSI0 (|Y = 0) =n=1
(1)n+1 n!
BA
n
(2.16)
with BA 2S .Eq. (2.14) can be rewritten in a more convenient form using the Cauchy formula for the derivatives, namely,
nZMFA(R|Y)nR
= n!1
2 i
C
ZMFA(R|Y)n+1R
d R; (2.17)
The contour C in Eq. (2.17) is a circle with a small radius around R = 0. However, since the function Z does not
grow at large R for n 1 we can close our contour C on the singularities of the function Z. We will call this newcontour CR.
NMPSI0 (Y) = 1 exp
BA
(1)R
(2)R
NMFA
(
(1)R |Y Y
NMFA
(
(2)R |Y
|
(1)R
= (2)R
=0
= 1 n=1
BAnn!
n! n! 1(2 i)2
C1R
d(1)RZMFA((1)R |Y Y)
((1)R )
n+1
C2R
d (2)RZMFA((2)R |Y)
((2)R )
n+1=
=1
(2 i)2
d
(1)R
(1)R
d (2)R
(2)R
1 exp
(1)R
(2)R
BAeY
(1)R
(1)R
BAeY
0,
(1)R
(2)R
BAeY
ZMFA
(1)R
ZMFA
(2)R
(2.18)
Here we introduce new variables (1)R =
(1)R exp(Y Y) and (2)R = (2)R exp(Y). In these new variables
ZMFA
(1)R
=
1
1 + (1)R
; ZMFA
(2)R
=
1
1 + (2)R
(2.19)
Closing the integration on the poles
(1)
R = 1 and (2)
R = 1 we obtain the formula of Eq. (2.15).
3. Calculation of the simplest diagrams in the BFKL Pomeron calculus
The main goal of this paper is to show that in the general case of the BFKL Pomeron calculus we have the same
situation as in the simple model. We start from the analysis of the simple diagrams of the BFKL Pomeron interaction.
6
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
8/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
9/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
10/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
11/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
12/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
13/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
14/36
transferred along the Pomeron are much smaller than momenta conjugate to dipole sizes 2 |k| |q|, which allows usto simply Eq. (3.32) using | k1 + q1/2 q2/2| |k
1| and | k1 q2/2| |k
1|. It is easy to see that the integration
over k1 leads to a delta function (i + 21 22 23), which is the conservation of the anomalous dimensions( = 1/2 i) in the triple Pomeron vertex (1 + 1 2 3) (see Ref. [15] where this -function was found for theforward scattering amplitude). However we need to be careful since in Eq. (3.32) we implicitly used the condition
that k23 < k21 < k
21. For k
21 > k
21 the second term in Eq. (A-21) gives the main contribution. Therefore we integrate
over all the kinematic region from k21 = 0 to k21 = but subtract the regions k21 > k21 and replace them by correct
integral.
For example for k21 > k21 we obtain the following integral
k21k23
i1 l
dl e(1+i212i22i3) l
k21k23
i1 l
dl e(1i212i22i3) l
(3.33)
where l = ln(k21/k23). The result of the integration leads to
e(1i2i3) l 1
1 + i21 i22 2i3 1
1 i21 i22 2i3
(3.34)
It is easy to see that the integral over 1 is equal to zero. The behaviour of (1) at large values of 1 we regularise
by adding a small i term for 1 ((1) (1 + i). (1 + i) 1i providing a good convergence ofthe integral on the large circle.
Therefore, finally we reproduce the (1 2 3) contribution.After further integration over q
1, q
2and d1 in Eq. (3.32) we obtain
A(P
2P; Fig. 4) =
22s
NcY0
dY d2
d3
1
2(2)2(2)(q
3 q1
+ q2
) (4)3(0, 2
) (0, 3
)
e(i/223)(YY) e(2)(Y
0) e(3)(Y0) |k1|12i212i3 |k2|1+2i2 |k3|1+2i3
=22s
Nc
d2 d3
1
2(2)2(2)(q
3 q
1+ q
2)
(4)3(0, 2) (0, 3)
(i/2 + 2 + 3) (2) (3) (3.35)
|k1|12i212i3 |k2|1+2i2 |k3|1+2i3
e(i/2+2+3)(Y0) e((2)+(3))(Y0)
As was already shown in the toy model in Eq. (2.1) the integration over rapidity brings in two terms in the last line
of Eq. (3.35). The first term corresponds to single Pomeron exchange and the second term corresponds to double
Pomeron exchange of two non-interacting Pomerons. To see this we rewrite the two terms in a more convenient form.
The first term in the last line of Eq. (3.35) reads
22sNc
12
(2)2(2)(q3 q
1+ q
2)
d2 d
(4)3(0, 2) (0, i/2 + 2)() (2) (i/2 + 2) (3.36)
|k1|12i |k2|1+2i2 |k3|1+2i12i2 e()(Y0)
2To see this It is enough to recall that we can safely use MFA for the nucleus target where k1 k2 1/R but q1 q2 1/RAwhere R and RA are nucleon and nucleus radii and RA R.
13
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
15/36
where = i/2 + 2+ 3. At high energies the leading contribution comes from small values of and we can simplifyEq. (3.36) as
22sNc
1
2(2)2(2)(q
3 q
1+ q
2)
d2
(4)3(0, 2) (0, 2 + i/2)(0)
(2)
(
2 + i/2)
1
k2k3 k3k2
2i2
d
1
k1k3 k3k1
2i
e()(Y0) =
22sNc
1
2(2)2(2)(q
3 q
1+ q
2)f(k2, k3) 2
2P(k3; k1|Y 0) (3.37)
where P(k3; k1|Y 0) is a single Pomeron exchange defined in Eq. (3.29) and f(k2, k3) is a function of k2 and k3which has to be calculated and is given by
f(k2, k3) =
d2
(4)3(0, 2) (0, 2 + i/2)(0) (2) (2 + i/2)
1
k2k3
k3k2
2i2(3.38)
The second term of Eq. (3.35) represents two non-interacting Pomeron exchange. To see this we recast it in the
form of
22s
Nc d
2d
3
1
2(2)2(2)(q
3 q1
+ q2
)(4)3(0, 2) (0, 3)
(i/2 + 2 + 3) (2) (3) (3.39)
1
k1k2
k2k1
2i2e(2)(Y0)
1
k1k3
k3k1
2i3e(3)(Y0) (3.40)
At high energies the main contribution comes from small values of2 and 3 and thus we can simplify Eq. (3.39) as
follows
22sNc
1
2(2)2(2)(q
3 q
1+ q
2) (42)2
(4)3
(i/2) (0) (0) (22)2P(k2, k1|Y 0) P(k3, k1|Y 0) (3.41)
Comparing Eq. (3.35) with Eq. (2.1) using the last result we conclude that in QCD the fan diagrams can be
viewed as the contribution from the exchange of two non-interacting BFKL Pomerons and a renormalisation of the
single Pomeron contribution.
However, as it was noticed by Hatta and Mueller [ 44], there is another saddle point in Eq. (3.35) where the
denominator (2 + 3 i/2) (2) (3) is close to zero. Indeed, the equation for this saddle point ( SP =1/2 iSP,1) looks as follows
2 (SP) Y 2((2 (SP 1) (SP)
(2 SP 1) 2 (SP) + ln(k21/k
23) = 0 (3.42)
In Eq. (3.42) we assume that the main contribution is dominated by 1 = 2 One can see that the position of this
saddle point is very close to the solution of the equation
(2 0 1) 2 (0) = 0 (3.43)
Denoting SP 0 by one can see from Eq. (3.42) that for the first term in Eq. (3.35) we have =
1
2 (0) Y + ln(k21/k23)
0 (3.44)
and for the second one
=1
2 (20 1) Y + ln(k21/k23) 0 (3.45)
14
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
16/36
2 ()
(2 1)
(2 1)
0
-10
-5
0
5
10
15
20
25
30
35
40
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
pert
(2 1)2sat
()
1.5
2
2.5
3
3.5
4
4.5
5
0.55 0.6 0.65 0.7 0.75 0.8
Figure 5: Solution to Eq. (3.43) for s, given by the
BFKL equations.
Figure 6: Solution to Eq. (3.43) for pert given by the
BFKL equation and sat calculated from Eq. (3.47).
The sum of these two terms leads to the contribution with energy dependence
A0 (P 2P; Fig. 1) e((0+i)+(0i)) (Y0)
It turns out (see Fig. 5) that the resulting intercept 2 (0) is larger than the intercept for the two Pomeron
exchange (2( = 1/2)) and, therefore, this contribution is the largest among the three. The appearance of such new
singularities in the angular momentum plane is a deadly blow to the entire approach based on the BFKL Pomeron.
Indeed, since the new singularity (double pole) is located to the right of the singularity generated by the exchange
of two BFKL Pomerons, we need firstly to sum over all such singularities, to obtain the resulting Green function
and only after doing this can we build the Reggeon calculus based on this Green function. However, the situation
changes crucially if we take into account the fact that in Eq. (3.43) (0) and (201) are actually in quite differentkinematic regions. (2 1) enters with a small value of the argument, namely 20 1 < cr and, therefore, canbe calculated using the BFKL equation of Eq. (3.7). We recall that cr can be found from the following equation
[1, 46, 47, 48] (cr)
1 cr = d (cr)
d cr(3.46)
However, we have a different situation for (0). Indeed, 0 turns out to be larger than cr (see Fig. 5 and recall
that cr = 0.37). Therefore, (0) describes the behaviour of the scattering amplitude in the saturation region, and,
therefore, cannot be calculated using Eq. (3.7) (see also Ref.[45], where the same conclusion has been derived from
slightly different considerations). As it was found in Ref. [46] for in the saturation region we have
sat() =(cr)
1 cr (1 ) (3.47)
at least for > cr
but close to cr
.
As one can see from Fig. 6 Eq. (3.43) which can be written as
2sat(0) = pert(20 1) (3.48)
has no solution.
15
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
17/36
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
18/36
reduces to the following contribution
A (Fig. 7) =
22SNc
22 d
2i=1
2i (i) 2 di d d2q d2k1 d2k2
g (k, 0, 0, ) g (k1, 0, 0, ) g(k1 + q/2, q, 0, 1) g(k1 q/2, q, 0, 2) g(k2 + q/2, q, 0, 1) g(k2 q/2, q, 0, 2) g (k4, 0, 0, ) g (k4, 0, 0, )
12i
a+iai
d eY1
()1
(1) (2)1
() (3.51)
The integral over can be rewritten as the sum over different contributions. We will show below that = andthe integral has the form
1
2i
a+iai
d eY1
( ())21
(1) (2) (3.52)
e()Y Y 1
()
(1)
(2)
1
()
(1)
(2)
+ e((1)+(2))Y1
(()
(1)
(2))2
The first term leads to the renormalisation of the BFKL Pomeron intercept (the first term in the brackets) as
well as of the vertex of the interaction of the BFKL Pomeron with the target (see Fig. 3). The second term reduces
to the exchange of two BFKL Pomerons without any interaction between them. However, for 0 =12 i0, given by
Eq. (3.43), instead of Eq. (3.52), we have the pole of the third order which leads to the contribution
1
2i
a+iai
d e Y1
( ())21
(1) (2) =1
2Y2 e(0) Y (3.53)
We will not consider this contribution by the same reason that we used in calculating the fan diagram since in this
diagram two Pomerons with and turn out to be outside of the saturation region while two Pomerons with 1 and2 are located inside the saturation region where we cannot use the BFKL kernel to determine the values of their
intercepts. Therefore, the first fan and enhanced diagrams are very similar with respect to the integration overrapidity. However, in the case of the enhanced diagram we have an additional problem: the q integration. Indeed,even if we assume that k > k0 and q = 0, the integration over q is restricted by the smallest of the two momenta k1and k2(see Appendix B for more details on q integration). In other words we have the following region of integration:
1. k > k1 > k2 > k0.
In this region the first term in Eq. (A-21) contributes for all four Pomerons and we have the product of two
functions in the vertices, namely, (1 + 1 2) (1 + 1 2). These leads to = ( = ) and theentire contribution looks the same as in the case of the first fan diagram. It is worthwhile mentioning that
for dipoles inside one BFKL Pomeron we have the same ordering in momenta.
2. k > k1
> k0
but k1
< k2
while k2
> k0
. In this kinematic region the first Pomeron with and the fourth
one with have contributions which stem from the first term in Eq. (A-21). However, for two Pomerons with1 and 2 the second terms in Eq. (A-21) play the most important roles. Collecting all factors of k1 one can
see that the integration over k1 has the formd l exp
(1 1
2 i i1 i2) l
(3.54)
17
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
19/36
where l = ln(k21/k20). The first term (1) in the bracket of Eq. (3.54) stemming from the integration over q1 k11,
the second term (12 i) originates from the contribution of the Pomeron with , the third as well as thefourth terms reflects the product of the factor (k21)
12+i1 from the upper vertex and factor (k21)2i1 from the
lower vertex since q21 = k21. The integration generates (2 1 2). The integration over k2 has the same
character as for the case of the first term and leads to (
1
2). These functions give 1
= .
The decomposition of Eq. (3.52) gives once more the terms shown in Fig. 3 and the pole of the third order
of Eq. (3.53) since (1 ) = () (see Eq. (3.7)). There is only one difference: in this pole of the thirdorder, three Pomerons (with , 1 and 2) are inside of the saturation region. This fact does not influence our
reasoning that we need not consider this kind of contribution.
Finally, we can conclude that the first enhanced diagram has the same three contributions as the BFKL Pomeron
calculus in zero transverse dimensions.
3.5 Reduction of the emhanced diagrams to the system of non-interacting Pomerons
In this section we will prove that an arbitrary enhanced diagram (see the diagram in Fig. 1-1) can be reduced to the
system of non-interacting Pomerons in the kinematic region: SY
1/S . As far as integration is concerned
this diagram can be written as a+iai
d
2ieY 2B
1
()m+1
m1 (1P I|) (3.55)
where B is the vertex of Pomeron interaction with the dipole3 (1P I|) is the one Pomeron irreducable diagram
and can be written in the following form
(1P I|) = 2
(1) (2) +j
(1)j jj1i>2
1
il,l>2 (l) (3.56)In Eq. (3.55) and Eq. (3.56) denotes the Pomeron - dipole vertex and is the triple Pomeron vertex (see Eq. (3.25)).
The integral over in Eq. (3.55) can be taken closing contour over one Pomeron poles = () ( see the diagramin Fig. 1-2) and closing contour over singularities of (1P I|) - the diagram in Fig. 1-3. The first contribution hasthe form
A(Fig. 1 2) 1m!
(Y + O (1/S))m
exp(()Y) m1 (1P I| = ()) (3.57)The contributions of the order of 1/S stem from the differentiation of the factor m1 (1P I|) with respect to inEq. (3.55) in taking the pole of m + 1-order. All these contributions are so small that they can be neglected, since
Y 1/S (SY 1).From Eq. (3.56) one can see that the first term in 1 (1P I|) is of the order of 2/() 3S while the others
lead to smaller contribution. Therefore, the only term with m = 0 remains that gives the exchange of one Pomeron
(see Fig. 1-4).
The first singularity that 1 (1P I
|) has is the pole
(1) + (2) . These poles are shown explicitly in
Fig. 8-3 and closing contour on these poles we obtain the contribution which is similar to Eq. ( 3.57), namely,
A (Fig. 8 5) 1m2!
N22 ( = (1) + (2)/ (Y + O (1/S))m2 1 (2P I| = (1 + (2)) (3.58)
3In this section we use for this vertex the notation B instead since we reserve for the anomalous dimension. We hope that in
other part of the paper we use for the vertex will not lead to any misunderstanding.
18
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
20/36
Figure 8: Reduction of the enhanced diagrams to a system of non-interecting Pomerons .
= (1) + (2)
= (1) + (2) + (3)
A)
B)
C) 2S
2S
= (1) + (2)
2S
Figure 9: The vertices of two and three Pomerons interaction with dipoles and (2PI| = (1) + (2)) .
where m2 is the number of the two Pomeron poles4; N2 is the vertex of dipole-two Pomeron interaction (see Fig. 9-
A) and 1 (2P I|) is the sum of two Pomeron irreducable diagrams. Since the first term in 1 (2P I|) gives thecontribution which is proportional to 3S (see Fig. 9-B) we can neglect all contribution except with m2 = 0 reducing
the diagram of Fig. 8-5 to two Pomerons exchange of Fig. 8-8. To take the contribution from three Pomerons exchange
in the diagram of Fig. 8-3 we need to consider the diagram of Fig. 8-6 and close the contour in over the poles
= (1
) + (2
) + (3
). Repeating the same procedure and taking into account that for the vertex of dipole-three
Pomerons we have the relation shown in Fig. 9-C we obtain the diagram of Fig. 8-9. Continuing this procedure we
see that we reduce the general enhanced diagram to the system of non-interacting Pomerons.
4We assume that 1 and 2 are the same for all two Pomerons poles. Actually, only sum of1 + 2 preserves. Strictly speaking we
need to consider all two Pomerons poles a bit different and take separately each of them. It is easy to see that in doing so we get the
same result as in Eq. (3.58) after integration over relevant s
19
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
21/36
4. General solution for the simplified BFKL kernel
4.1 The main idea
The analysis of the first diagrams for the BFKL Pomeron calculus shows that they have the same kind of contributions
as for the BFKL calculus in zero transverse dimensions and the new one that stems from the possibility that for
some specific values of the anomalous dimension (0,k) the intercept of n BFKL Pomerons in exchange is equal to
the intercept of k BFKL Pomerons with k < n. However, for this specific contribution two or more Pomerons are
located inside the saturation region where we cannot use Eq. ( 3.7) to calculate the intercept of the Pomeron. Inside
the saturation region we can use Eq. (3.47) which has a rather general proof (see Ref. [46]). In doing so we see that
this specific QCD contribution leads to an energy suppressed scattering amplitude and can be neglected in the first
approximation to the problem.
Having this result in mind we can restrict ourselves by calculating only terms stemming from the decomposition
of the type of Eq. (3.52). This decomposition leads to the exchange of non-interacting BFKL Pomerons and to the
renormalisation of the vertex for the Pomeron target (projectile) interaction as well as to the renormalisation of the
Pomeron intercept. In the kinematic region given by Eq. (1.1) we can neglect the renormalisation of the Pomeron
intercept and find the renormalised vertex for the Pomeron target (projectile) interaction from the initial conditionto our problem since it is closely related to the dipole -dipole (or target) interaction at low energies.
Since we are dealing with the system of non-interacting Pomerons we can apply the procedure of the MPSI
approximation (see Ref. [23]) to take into account all the Pomeron loops. The strategy of the improved MPSI
approach consists of three major steps
1. To find the solution in the mean field approximation for arbitrary initial conditions. According to our analysis
in this approximation the scattering amplitude has the following form
N(Y Y, [R(ki, bi)]) =
n=1 n
i=1d2ki (1)n Cn(k) P (k, ki; bi|Y Y) R(ki, bi) (4.1)
where R is an arbitrary function of k (recall that k is the conjugated variable to the size of a dipole).
2. Comparing the exact solution with the initial condition to find the solution in terms of the renormalized
vertices. It means that we need to solve the functional equation (in the general case)
N(Y Y = 0, [R(k, b)]; b) = N([(k, b)]) (4.2)
where the function N should be given. In other words the scattering amplitude for low energies should be
known either from calculations in QCD or from some phenomenology. This step is in full agreement with the
parton model where the high energy interaction can be calculated through the low energy amplitude for the
interaction of wee partons that have to be given.
3. To use the MPSI formula to take into account the Pomeron loops. This formula reads as follows
NIMPSI(k, k0; b, Y 0) =n=1
(1)n1n!
ni=1,j=1
d2kid2kjd
2bid2bj
(1)R (ki, bi)
(2)R (kj , bj)
NMFA
Y Y, [(1)R (ki, bi)]
NMFA
Y, [(2)R (kj , bj)]
|(1)R
=(2)R
=0BA
ki, kj , b bi bk
(4.3)
20
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
22/36
where BA is the amplitude in the Born approximation. It should be stressed that Eq. (4.1) for NMFA
guarantees that NIMPSI(k, k0; b, Y 0) does not depend on the value ofY. Indeed, using the fact thatd2ki d
2 kj d2bi d
2bjP(k, ki; bi|Y Y) BA
ki, kj , b bi bk
P (k0, kj ; bj|Y) = 2S P(k, k0; b|Y) (4.4)
one can rewrite Eq. (4.3) in the form
NIMPSI(k, k0; b, Y 0) =n=1
(1)n1n!
ni=1,j=1
d2kid2kj d
2bid2bj
(1)R (ki, bi)
(2)R (kj , bj)
NMFA
Y Y, [(1)R (ki, bi)]
NMFA
Y, [(2)R (kj , bj)]
|
(1)R
=(2)R
=0
2S P (k, k0; b|Y)
n= 1 exp
2S P(k, k0; b|Y)
(1)P
(2)P
NMFA
(1)P
NMFA
(2)P
|(1)P
=(2)P
=0(4.5)
where (1)P = P (k, ki; bi|Y Y) R(ki, bi) and (2)P = P (k0, ki; bi|Y) R(ki, bi)
At the moment we can carry out this program only for a simplified BFKL kernel since only for this kernel do wehave the exact analytical solution to work with (see [6]). The substantial number of numerical solutions [8] does not
help us since we know how to perform the renormalisation of the vertices only for the analytical solution.
4.2 The simplified BFKL kernel and MFA solution
The kernel for which we find the solution and implement our program has the following form [6]
( =1
2+ i) = S
1 for r
2 Q2s 1 ;1
1 for r2 Q2s 1 ;
(4.6)
This kernel sums the leading log contributions in the region of low x. The first one is the usual LLA approximation
of perturbative QCD in which the terms of the order of S
ln(r2 2)n are taken into account for r2 Q2s
1
where S ln(r2 2) 1; and the second log approximation leads to the summation of the terms of the order of
S ln(r2 Q2s)n
in the kinematic region where r2 Q2s 1 and S ln(r2 Q2s) 1 [49, 6].The LLA leads to the ordering of the dipole momenta.
In the perturbative QCD region (r2 Q2s 1) we have, using the diagrams of Fig. 7 as an example,
k k1 k2 k0 (4.7)
In the saturation region the ordering is the opposite, namely,
k k1 k2 max( k0 or Qs) (4.8)
where Qs is the saturation scale.Eq. (4.6) is not only much simpler than the exact kernel of Eq. ( 3.7) but calculating the Pomeron diagrams in
LLA we can neglect the contribution due to the overlapping of the intercept for a different number of Pomerons in
exchange. Indeed, the fact that a number of Pomerons are in the saturation region while other Pomerons are outside
this region, means that we cannot keep the ordering of Eq. ( 4.7) or Eq. (4.8). Therefore, dealing with this kernel we
enhance the arguments for neglecting the overlapping singularities.
21
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
23/36
Fortunately, the solution for the MFA has been found in Ref.[ 6] in the entire kinematic region for arbitrary initial
conditions. For completeness of our presentation, we will describe the main features of this solution inside of the
saturation region in this section. The solution is written for the amplitude in the coordinate representation, namely
N(x12; b; Y) = x212 d
2k
(2)2eikx12 N(k, b; Y) (4.9)
The solution in the MFA ( to the Balitsky-Kovchegov equation) can be written in the form
N(z) = 1 e(z) (4.10)where the variable
z = ln
x212 Q2s(Y, b)
(4.11)
and function (z) is determined by the following equation
z =
2
0(b)
d
+ (exp() 1) (4.12)
The boundary conditions for the solution of Eq. (4.10) we should find from the solution at z < 0. We assumethat to the right of the critical line x212 Q
2s(Y, b) 1 (z = 0) the exchange of one BFKL Pomeron gives the main
contribution. It has the form [7] for our simplified BFKL kernel
P (z, ) = e12 z eb
2/R2 = (b) e12 z (4.13)
where R2 ( in the spirit of the LLA approximation ) depends only on initial conditions at low energy and has a clear
non-perturbative origin. Therefore the boundary conditions have the form
N(z = 0+; Eq. (4.10)) = N(z = 0) = eb2/R2 (b) ; (4.14)
d ln N
dz|z=0+ =
d ln N
dz|z=0 =
1
2; (4.15)
For small the solution to Eq. (4.12) has the form
ln
0(b)=
1
2z or = 0(b) e
12 z (4.16)
This equation says that we can satisfy Eq. (4.14) and Eq. (4.15) if
0(b) = (b) and (b) 1 (4.17)Since (b) 2S Eq. (4.17) shows that Eq. (4.12) gives the solution to our problem.
From Eq. (4.12) one can see that at small values of 0 the integral has a logarithmic divergence. Therefore, we
can rewrite Eq. (4.12) as follows
z + 2 ln (b) = 2 ln (b) e 12 z = 2
0(b)
d + (exp() 1)
+ ln 0(b) (4.18)
The r.h.s. of this equation does not depend on 0(b) = (b) (see Eq. (4.17)). In Fig. 10 we plot (z) for two cases :
the exact expression of Eq. (4.12) and
2
a
d + (exp() 1) + ln (0(b)/a) (4.19)
22
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
24/36
One can see that both functions coincide. This is an illustration that the integral of Eq. (4.12) has no further
dependence on (b). Having this in mind we can claim that Eq. ( 4.18) leads to a general function, namely
(z) =
(b) e
12 z
(P(z, b)) (4.20)
The argument of the function is the one Pomeron exchange (see Eq. (4.13)) and expanding
(b) e12 z
in a series
with respect to this argument we are able to find the coefficient Cn in Eq. (4.1).
Eq. (4.12) has also two simple analytical solutions for the function : in the region of small ( see Eq. (4.16)
and for large . The last solution is
(z) =1
2ln2
(b) e
12 z
(4.21)
The exact solution is shown in Fig. 11 together with these two analytical solutions. One can see that these two
analytical solutions give a good approximation for small and large values of (b) e12 z.
z()
exact
approximate
6
8
10
12
14
16
2 4 6 8 10
(z)=0.01
exactapproximate
small z
large z
z
0
2
4
6
8
10
12
14
0 2 4 6 8 10 12 14
Figure 10: Comparison of the exact integral of
Eq. (4.12) with the approximate formula of Eq. (4.19)
with a = 0.25.
Figure 11: Solution to Eq. (4.12) (solid line) and two
approximate analytical function: Eq. (4.16) (small z) and
Eq. (4.21) at large z (both are shown in dotted lines).
4.3 Improved MPSI solution for Pomeron loops
Using Eq. (4.10) and Eq. (4.20) we can find the sum of the Pomeron loops by means of Eq. (4.5). We demonstrate the
answer by solving analytically two extreme cases: (i) the behaviour of the scattering amplitude near to the saturation
scale; and (ii) the asymptotic behaviour deep inside of the saturation region.
For the first case = (b)exp12
z
and Eq. (4.10) together with Eq. (4.5) lead to the following answer
NIMPSIx2, R2; Y = 1 exp (b) Px2; R2; Y = 1 exp(b) e
12z (4.22)
where x and R are the sizes of the projectile and target dipoles and z is defined by Eq. (4.11) where Q2s 1/R2.For finding the behaviour of the scattering amplitude deeply inside of the saturation region we will use Eq. (2.18).
In new variables: 12 l l1 = ln
(1)P
and 12 l + l1 = ln
(2)P
, the contour C looks as it is shown in Fig. 12. Since
NMFA does not have singularities in the variables l and l1 we can replace the contour C by the contour CR (see
23
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
25/36
Fig. 12). Therefore, we can rewrite Eq. (2.18) in the following form
NMPSI0 (z) = 1 1
(2 i)2
CR
d l
CR
d l1 exp
el
P (BA, z)
el
P (BA, z)
0,
el
P (BA, z)
ZMFA Pe 12 l+ l1 ZMFA Pe 12 ll1 (4.23)
Using Eq. (4.13) and Eq. (4.21) we rewrite
ZMFA
P
e12 l+ l1
= 1 NNFA
(1)P
= e(
12 l+l1) = e
12 (
12 l + l1)
2
(4.24)
ZMFA
P
e12 l l1
= 1 NNFA
(2)P
= e(
12 ll1) = e
12 (
12 l l1)
2
where z = z1 + z2.
One can see that the integration over l1 reduces to a very simple integral, namely
1
2i
CR
d l1 el21 =
1
e
2
0
d l1 sin (2 l1) el21 =
1
2
erfi() (4.25)
where erfi(z) = erf(i z)/i and erf(z) E2(z) = 2z0
, exp t2 dt is the error function (see Refs. [40, 41]) .
The remaining integral over l has the form
NMPSI0 (z) = 1 (4.26)
erfi()2
0
d l
1 exp
1
BA e12zl
1
BA e12 zl
0,
1
BA e12zl
sin(2l) e 18 l2
It is clear from Eq. (4.26) that the dominant contribution stems from l of the order of unity due to the exponential
decrease at large l. Therefore, the behaviour of the integrand at large exp (12z) will dictate the resulting approach
to 1 for the amplitude. At z 1 the integrand gives (b) exp( 12z + l) behaviour. We can take the integral over lwhich leads to the following answer:
erfi()
2
0
d l
(b) e12zl
sin(l/2) e 18 l2 = (4.27)
erfi()
2
ie2
2
erf
4 + i
2
2
+ i erfi
4 + i
2
2
1
(b) e12 z
The solution of Eq. (4.27) shows the geometrical scaling behaviour and at high energies N 1 exp 12z.In such a behaviour two features look unexpected: the geometrical scaling behaviour since the statistical physics
motivated approach leads to a violation of this kind of behaviour [ 17]; and the fact that the asymptotic behaviour
shows a very slow fall down in comparison with the expected N 1 exp C z2 [6, 23, 36]. This slow decreaseshows that the Pomeron loops essentially change the behaviour of the amplitude in the saturation region.
Our main difference to the statistical physics motivated approach is the fact that we took into account the
impact parameter behaviour of the scattering amplitude. We observed that the integration in the Pomeron loops isdominated by the impact parameters of the order of the largest dipole size involved in two vertices assigned to the
loop. We can face two different possibilities. In the first case the largest dipole belongs to the vertex , for which we
assign our rapidity and for which the generating functional or other equation for our statistical system is written.
This is the lower vertex of the Pomeron loop at the lower value of rapidity. In this case everything is in agreement
with the statistical approach. In the second case the largest dipole belongs to the vertex at the top of the Pomeron
24
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
26/36
loop at higher rapidity value which is our future in terms of the statistical approach. The future behavior of this
situation cannot be described by a Markov chain without taking into account the long range rapidity correlations.
It should be stressed that the appearance of the overlapping singularities cannot be included in the statistically
motivated formalism until the problem is reformulated in terms of the interaction of new effective particles related
to these singularities.
4.4 Self-consistency checkCR
C
i
i
l
Figure 12: The contours of integration in Eq. (4.23).
As has been discussed we need to come back and consider
the overlapping singularities that have been neglected. The
check of self-consistency of our approximation means that
we need to calculate all diagrams but replace the exchange
of the BFKL Pomerons by the sum of all enhanced dia-
grams (see Fig. 13) in the kinematic region where the BFKL Pomeron enters inside of the saturation region. For
example, the simplest diagram of Fig. 7 has to be replaced by the diagram of Fig. 14 in the kinematic region that is
related to the overlapping singularities.
For the simplified kernel the value of 0 (see Fig. 6) is equal to 0.6 and this value is close to cr = 1/2. However,since 0 cr does not depend on the value of rapidity, at high energies the solution that we need to check will beat z 1.
Indeed, the solution of Eq. (3.28) has the form of Eq. (4.13) only at small values of . In the region of small
the expression of Eq. (3.28) has the form
P(k, k0|Y) = 1k2 k20
d e()Y
k2
k20
i 1
d exp
1
2z iz 8 2 SY
exp+ 12 z z2
32 S Y
(4.28)
In the last line of Eq. (4.28) we used the steepest decent method with the saddle point value of
= SP = i z16SY
(4.29)
Therefore, Eq. (4.13) for the Pomeron holds only for z 4 SY and at SP 1. In this kinematic regionSP 1. In other words, the region = 12 i 0 corresponds to the large values of z where our scatteringamplitude behaves as
A 1 e 12 z (4.30)
For the Pomeron near and inside of the saturation region we can rewrite the Pomeron diagrams in terms of the
variables z and k introducing the Mellin transform for the Pomeron in the form
P (z) =1
2i
a+iai
de z
(4.31)
where the intercept 0 =12 .
25
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
27/36
= + + = Sum of all enhanced diagrams
Y1, k1
Y0 = 0, k0
(2),2,qq1(1),1,q
Y2,k2
Y, k
(), , q
(),, q
Figure 13: The graphic form for notation of wave bold
line : the sum of all enhanced diagrams.
Figure 14: The diagram that we need to replace the one
in Fig. 7 to study the overlapping singularities. By bold
line we denote the sum of diagrams of Fig. 13.
Eq. (4.30) has the form in -representation:
A =1
C
+ 0(4.32)
the coefficient C in Eq. (4.32) reflects the fact that we know the asymptotic behaviour inside the saturation region
within the pre-exponential accuracy.
In this representation the diagram of Fig. 14 will be proportional to
A (Fig. 14) 12i
a+iai
d ez1
0
1
C
+ 0
1
0 (4.33)
instead of Eq. (3.51).
One can see that we do not have the overlapping singularities. The main contribution stems from = 0 and,
since we have a double pole for this value of the diagram describes the renormalisation of the Pomeron intercept
which is proportional to 3S and can be neglected in the kinematic region of Eq. ( 1.1).
5. Conclusions
The main results of this paper are:
1. In the kinematic region of Eq. (1.1) the BFKL Pomeron calculus after renormalisation of the Pomeron - target
vertex, can be reduced to summing over the non-enhanced diagrams that describe the system of non-interacting
Pomerons;
2. The sum of enhanced diagrams for the system of non-interacting Pomerons can be calculated by means of the
improved Mueller-Patel-Salam-Iancu approach, using an additional piece of information on dipole scattering
amplitude at low energy (for example, using the Born Approximation of perturbative QCD for its determina-
tion);
3. For the simplified BFKL kernel given by Eq. (4.6) the analytical solution was found to have two unexpected
features: the geometrical scaling behaviour is deeply in the saturation region and is slow to approach (N 1 Cexp(12z) instead of N 1 Cexp(z2/8) as it happens in the MFA formalism.
26
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
28/36
In this paper we confirm the result of Hatta and Mueller [44] that a new type of singularities appear in the
BFKL Pomeron calculus: the overlapping singularities. They lead to a steeper asymptotic behaviour than the multi
BFKL Pomeron exchanges give. It means that even the MFA formalism is doubtful without talking about summing
over Pomeron loops. We argue that we can neglect these singularities since they stem from the kinematic region
where two or more BFKL Pomerons are in the saturation region. We did the self-consistency check that such kind
of singularities do not appear in the framework of the solution that we found.
As has been mentioned several times in this paper, the geometrical scaling behaviour is in contradiction with
the statistical physics analogy description and the Langevin equation based on this analogy[ 17]. However, we need
to recall that the statistical-like approach at the moment exists only as a QCD motivated model since the form of
the noise term that has been used, is oversimplified. Indeed, we can prove that the BFKL Pomeron calculus can
be described by the Langevin equation with some noise term but this noise is so complicated ( see [50]) that we do
not know how to treat it. Instead of this the simplified model for the noise term has been used. This model, as
far as we understand it, has its own difficulties: (i) it cannot describe the overlapping singularities that have been
discussed here; (ii) it did not include the impact parameter dependence which is very important for any treatment
of the Pomeron loops as it was shown here; and (iii) even for the toy model in which we neglect the dependence of
the vertices on the dipole sizes, the statistical model predicts the constant cross section at high energies 5 while the
exact solution (see Ref. [28] ) leads to a decrease of the total cross section.
We believe that the most vital problem, among these, is the correct impact parametre dependence. The calcu-
lation of the enhanced diagrams show that the typical impact parameter in the Pomeron loops is of the order of the
smallest size of the dipoles involved in the two vertices in the loops. It might or might not be the dipole from the
vertex at rapidity Y for which we write the equation. If this dipole belongs to the second vertex, it is related to how
we proceed from here and we do not think that it is possible to describe such a system in a statistical way, assuming
only short range correlations in rapidity. The failure of the toy model description by the Langevin equation is , in our
opinion, due to the fact that a statistical model cannot describe the renormalization of the intercept of the Pomeron
and their applicability is restricted to the kinematic region of Eq. (1.1).
Since we consider the case of fixed coupling , even in deep inelastic scttering we have to take into acount all
Pomeron loops. Indeed, as it was shown in Ref. [1] only running QCD coupling selects the fan diagrams and leads
to the mean field approximation for these processes. One cn see that the sum of Pomeron loopsat Q2 near to thesaturation scale has a typical eikonal form (see Eq. ( 4.22)). This result supports the numerous saturation models
that describes well the experimental data [53].
However, it seems more realistically, for a description of the deep inelastic processes first to solve the problem of
summation of the Pomeron loops in the case of running QCD coupling. In this case we expect [ 1] the the mean field
approximation will be able to describe the sunstantial part of kinematic region. More that ten years ago Braun and
Levin[54] suggested the procedure how to include the running QCD coupling in to th BKL equayion that describes
the linear evolution in the region of high energy. This suggestion was based on the strong assumption (which by the
way, were proved ) that the gluon reggeization being the essential part of the BFKL approach in the leading order
will be preserved in the next-to-leading order as well. The formula that they obtained has a so called triumvirate
structure, namely, for the basic process G(q) + G(q) G(q q) the fixed coupling constand should be replaced by
S S(q q) S(q)S(q)
Recently this conjucture has been proven (see Ref. [55]. This breakthrough allows us to include the running QCD
coupling in the equation of the high density QCD. However, an additional problem arise that should be studied,
5We got this information from S. Munier (lecture at GGI workshop on high density QCD) and from E. Naftali (private communication).
27
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
29/36
namely, such way of taking into account the running coupling effect induces the contriutions of infrared renormalons
(see paper of Levin in Ref. [54] ). Such a contribution cannot be treated perturbatively and can restrict the theoretical
accuracy of our approach. Therefore, summing Pomeron loops with running QCD coupling is our next challenging
problem.
Acknowledgments
We want to thank Asher Gotsman, Alex Kovner, Uri Maor, Larry McLerran and Al Mueller for very useful discussions
on the subject of this paper. Special thanks from A.P. goes to Carlos Pajares for his hospitality and support in the
University of Santiago de Compostela where this work was completed.
This research was supported in part by the Israel Science Foundation, founded by the Israeli Academy of Science
and Humanities, by a grant from Ministry of Science, Culture & Sport, Israel & the Russian Foundation for Basic
research of the Russian Federation, and by BSF grant # 20004019.
We thank the Galileo Galilei Institute for Theoretical Physics for the hospitality and the INFN for partial support
during the completion of this work.
Appendix A: Calculation of gk,q,n,
.
We want to calculate the Fourier transform of the vertex function E defined in Eq. (3.13) as
g(k, q, n, )
d2x12x212
d2z eikx12 eiqz E(x1, x2; x0|n; ) (A-1)
with z = (x1 + x2)/2 x0. The similar Fourier transform was already calculated in [52], but the authors used a
different normalization and we found it useful to calculate it using our notation. The integration over z was doneby Navelet and Peschanski in [43] by considering the solutions of the differential equations obeyed by the vertex
functions E and consequent matching of the result to an approximate one obtained by Lipatov in [42]. We adopt the
notation introduced by Lipatov for the mixed representation of the vertex functions
En,q () =22
bn,
1
||
dzdzei
2 (qz+qz)En,(z + /2, z /2) (A-2)
where
bn, =24i3
|n|/2 i(|n|/2 i+ 1/2)(|n|/2 + i)(|n|/2 + i+ 1/2)(|n|/2 i) (A-3)
and = x12 in our notation introduced in Eq. (3.6). The expression for E
n,
q () is given in [43] and reads
En,q () = qin/2q i+n/226i(1 i+ |n|/2)(1 i |n|/2)
Jn/2i(q
4)Jn/2i(
q
4) (1)nJn/2+i( q
4)Jn/2+i(
q
4)
(A-4)
Thus in terms of En,q () the Fourier transform in Eq. (A-1) is given by
28
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
30/36
g(k, q, n, ) =bn,22
d2
|| eik En,q () (A-5)
From Eq. (A-4) and Eq. (A-5) one can see that the calculation ofg(k, q, n, ) is reduced to evaluation of the following
integral
I,(q, k)
d2
|| eik J(
q
4)J(
q
4) (A-6)
We introduce polar coordinates for complex variables
q = |q|ei = ||ei k = |k|ei (A-7)
and rewrite Eq. (A-6) as
I,(q, k) 0
d
20
d eik cos() J(q
4ei())J(
q
4ei()) (A-8)
For the sake of simplicity in Eq. (A-8) and further below we denote by , q and k their respective absolute values.
We use the series representation for the Bessel functions
J(z) =z
2
m=0
(1)mm! ( + m + 1)
z2
2m(A-9)
to recast Eq. (A-8) into
I,(q, k) =0
d
20
d eik cos()
ei()
q
8
m=0(1)m
m! ( + m + 1) q
8 2m
ei2m() (A-10)
ei()
q8
m=0
(1)mm! ( + m + 1)
q8
2mei2m
()
Changing variables and using the integral representation of the Bessel functions
Jn(z) =1
2
ein+iz sin d =1
2
/2/2
ein(/2)+iz sin(/2)d( /2) (A-11)
=(i)n
2
/23/2
ein+iz cos d =(i)n
2
20
ein+iz cos d
we rewrite Eq. (A-10) as
I,(q, k) = m=0
m=0
0
d (2) J2m++2m(k)
iei()q
8
1m! ( + m + 1)
q8
2mei2m() (A-12)
i
ei()
q8
1m! ( + m + 1)
q8
2mei2m
()
29
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
31/36
Now we perform the integration over using the formula0
d J(k)q
8
=
1
k
q4k
(12 + +2 )(12 +
2 )
(A-13)
then the expression in Eq. (A-12) reads
I,(q, k) = 2k
iei() q
4k
m=0
1
m! ( + m + 1)
q4k
2mei2m()
1
(1/2 2m) (A-14)
i
ei()
q4k
m=0
1
m! ( + m + 1)
q4k
2mei2m
()(1/2 + + 2m)
One should note the full factorization of two series in Eq. (A-14) due to the integration over . This allows us to
sum them separately using
m=0
z2m
m! ( + m + 1)
1
(1/2 2m) =cos()
(1/2 + )
(1 + )2F1[1/4 + /2, 3/4 + /2 ; 1 + ; 4 z
2] (A-15)
and
m=0
z2mm! ( + m + 1)
(1/2 + + 2m) = (1/2 + )(1 + ) 2
F1[1/4 + /2, 3/4 + /2 ; 1 + ; 4 z2] (A-16)
With help of Eq. (A-15) and Eq. (A-16) we obtain the final expression for I,(q, k) as
I,(q, k) = 2k
iei() q
4k
cos()
(1/2 + )
(1 + )2F1[1/4 + /2, 3/4 + /2 ; 1 + ; 4
q4k
2ei2()]
i
ei()
q4k
(1/2 + )(1 + ) 2
F1[1/4 + /2, 3/4 + /2 ; 1 + ; 4
q4k
2ei2()] (A-17)
At this point we return to complex vector notation given in Eq. (A-7) and absorb the angles into complex vectors q,
k, q and k
I,(q, k) =
2
|k|i q
4k 1
(1 + )(1/2 )2F1[1/4 + /2, 3/4 + /2 ; 1 + ; q
2k2
]
(A-18)
i
q4k
(1/2 + )(1 + ) 2
F1[1/4 + /2, 3/4 + /2 ; 1 + ;
q2k
2]
In the first line in Eq. (A-18) we used identity (1/2 + z)(1/2 z) = cos(z)/. Plugging Eq. (A-18) into Eq. (A-5)we find the Fourier transform of the vertex functions E defined in Eq. (A-1)
g(k, q, n, ) =bn,22
in q in/2q i+n/226i(1 i+ |n|/2)(1 i |n|/2) In/2i,n/2i(q, k) (1)nIn/2+i,n/2+i(q, k) (A-19)with bn, given in Eq. (A-3). In the high energy limit the main contribution to Pomeron Green function comes from
n = 0. We are interested in the case of small momentum transferred along the Pomeron which implies
|k
| |q
|(A-20)
In the limit of Eq. (A-20) for n = 0 Eq. (A-19) reads
g(k,q, 0, ) =b0,22
|q| 2i26i2(1 i) cos(i) [Ii,i(q, k) I+i,+i(q, k)] = (A-21)
2
i22i |k|1+2i
2(1/2 i)(i)2(1/2 + i)(i)
2
i26i |k|12i |q|4i
2(1 i)(i)2(1 + i)(i)
30
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
32/36
Appendix B: Triple Pomeron vertex in q and representation.
In this appendix we calculate the triple Pomeron vertex in q and representation, namely,
3P (q1, q2; 1, 2, 3) d2k1 g k1, q1, 0, 1 gk1 +1
2 q2, q2, 0, 2
g
k
1 q1 +
1
2 q2, q1 q2, 0, 3 (B-1)where g is given by Eq. (A-21).
Introducing new holomorphic variables
t =q2
2k1 + q2
and t =q2
2 k1 + q2
(B-2)
and using Eq. (A-18) , we can rewrite Eq. (B-1) in the form
3P (0, q2; , 2, 3) = 4
q2
2
4
1/2i1+i2+i3 dtdt ((1 t) (1 t))1/2+i1 C1(1) (B-3)
C1 (2)2F1 1/4 + i2/2, 3/4 + i2/2, 1 + i2, t2 2F1 1/4 + i2/2, 3/4 + i2/2, 1 + i2, t2 C2 (2) (4 t t)2i2 2F1
1/4 + i2/2, 3/4 + i2/2, 1 + i2, t
2
2F1
1/4 + i2/2, 3/4 + i2/2, 1 + i2, t2
C1 (3)2F1 1/4 + i3/2, 3/4 + i3/2, 1 + i3, t2 2F1 1/4 + i3/2, 3/4 + i3/2, 1 + i3, t2 C2 (3) (4 t t)2i3 2F1
1/4 + i3/2, 3/4 + i3/2, 1 + i3, t
2
2F1
1/4 + i3/2, 3/4 + i3/2, 1 + i3, t2
with
C1 () =2
i 22i
2 (1/2 i) (i)2 (1/2 + i) (i) ; C2 () =
2
i 26i
2 (1 i) (i)2 (1 + i) (i) ; (B-4)
For simplicity we consider 3P (q1, q
2; 1, 2, 3) in Eq. (B-3) for q
1 = 0.
The integral of Eq. (B-4) has three region of potential divergency: t 0, t 1 and t . First, let us studythe behaviour of the integrant at t . For doing this we use ( see formula 9.132(2) of Ref. [40]) the followingexpression
2F1
1/4 + i2/2, 3/4 + i2/2, 1 + i2, t2
=(1 + i2)(1/2)
2(3/4 + i2)
1
t2
14+i
22
2F1
1/4 + i2/2, 1/4 + i2/2,
1
t2
(B-5)
+(1 + i2) (1/2)
2(1/4 + i2/2)
1
t2
34+i
22
2F1
3/4 + i2/2, 3/4 + i2/2,
1
t2
From Eq. (B-5) we conclude that the integrant of Eq. (B-3) decreases at large t as t2 for 2 = 3 = 0. It meansthat integrals over t and t are convergent as far as large t and t are concerned. The next suspicious region is t 0(t
0). Factor (t t)1/2+i1i2i3 in the integrant indicates a possible divergence in this region. Rewriting the
integral in a general form
3P (0, q2; 1, 2, 3) =
0
d(tt) (t t)1/2+i1i2i3 (t, t) (B-6)
=
10
d(tt) (t t)1/2+i1i2i3 {(0) + ((t, t) (0))} +
1
d(tt) (t t)1/2+i1i2i3 (t, t)
31
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
33/36
we see that if (t, t) (0) t t at t 0 we have a pole 1/(1/2 + i1 i2 i3). However, since our function(t, t) has factors t1/2+ii we need to be more careful. In calculating first enhanced diagram (see section 3.4) 2 and
3 are small and (t, t) has the following form
(t,t) = C1(1)
4
2 3 1 (4tt)2i2 1 (4tt)2i320, 30
C1(1) 44
ln
2
(4tt) (B-7)
Using Eq. (B-7) we obtain that
3P (0, q2; 1, 2, 3)
20, 30C1(1) 4
4q2
2
4
1/2i1+i2+i3(1/2 + i1 i2 i3)3 + function without singularities in 2 and 3
(B-8)
At tt 1 we could have a divergence which leads to a pole 1 /(3/2 + i1). However, since C1(1) has a zeroof the second order at i1 = 3/2 3P (0, q2; 1, 2, 3) has no such singularity. Therefore, Eq. (B-7) gives the triplePomeron at small 2 and 3. One can see that the pole in Eq. (B-8) replaces the -function which we obtained
integrating first of q2.
For arbitrary 2 and 3 the triple Pomeron vertex has a more complicated form, namely,
3P (0, q2; 1, 2, 3) = C1(1) 4
q2
2
4
1/2i1+i2+i3
C1 (2) C
1 (3)
1/2 + i1 i2 i3 C1 (2) C
2 (3)
1/2 + i1 i2 + i3(B-9)
C2 (2) C
1 (3)
1/2 + i1 + i2 i3 +C2 (2) C
2 (3)
1/2 + i1 + i2 + i3
+ regular function of2 and 3
Integration over q2 leads ( ) where (2 in Eq. (3.49)) corresponds to upper BFKL Pomeron in Fig. 7,and (4 in Eq. (3.49)) corresponds to lower BFKL Pomeron in Fig. 7.
References
[1] L. V. Gribov, E. M. Levin and M. G. Ryskin, Phys. Rep. 100, 1 (1983).
[2] A. H. Mueller and J. Qiu, Nucl. Phys.,427 B 268 (1986) .
[3] L. McLerran and R. Venugopalan, Phys. Rev. D 49,2233, 3352 (1994); D 50,2225 (1994); D 53,458 (1996); D 59,09400
(1999).
[4] I. Balitsky, [arXiv:hep-ph/9509348]; Phys. Rev. D60, 014020 (1999) [arXiv:hep-ph/9812311] Y. V. Kovchegov, Phys.
Rev. D60, 034008 (1999), [arXiv:hep-ph/9901281].
[5] J. Jalilian-Marian, A. Kovner, A. Leonidov and H. Weigert, Phys. Rev. D59, 014014 (1999), [arXiv:hep-ph/9706377];
Nucl. Phys. B504, 415 (1997), [arXiv:hep-ph/9701284]; J. Jalilian-Marian, A. Kovner and H. Weigert, Phys. Rev.
D59, 014015 (1999), [arXiv:hep-ph/9709432]; A. Kovner, J. G. Milhano and H. Weigert, Phys. Rev. D62, 114005
(2000), [arXiv:hep-ph/0004014] ; E. Iancu, A. Leonidov and L. D. McLerran, Phys. Lett. B510, 133 (2001);
[arXiv:hep-ph/0102009]; Nucl. Phys. A692, 583 (2001), [arXiv:hep-ph/0011241]; E. Ferreiro, E. Iancu, A. Leonidovand L. McLerran, Nucl. Phys. A703, 489 (2002), [arXiv:hep-ph/0109115]; H. Weigert, Nucl. Phys. A703, 823 (2002),
[arXiv:hep-ph/0004044].
[6] E. Levin and K. Tuchin, Nucl. Phys. A693 (2001) 787 [arXiv:hep-ph/0101275]; A691 (2001) 779
[arXiv:hep-ph/0012167]; B573 (2000) 833 [arXiv:hep-ph/9908317].
[7] E. Iancu, K. Itakura and L. McLerran, Nucl. Phys. A708 (2002) 327 [arXiv:hep-ph/0203137].
32
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
34/36
[8] N. Armesto and M. A. Braun, Eur. Phys. J. C20, 517 (2001) [arXiv:hep-ph/0104038]; M. Lublinsky, Eur. Phys. J.
C21, 513 (2001) [arXiv:hep-ph/0106112]; E. Levin and M. Lublinsky, Nucl. Phys. A712, 95 (2002)
[arXiv:hep-ph/0207374]; Nucl. Phys. A712, 95 (2002) [arXiv:hep-ph/0207374]; Eur. Phys. J. C22, 647 (2002)
[arXiv:hep-ph/0108239]; M. Lublinsky, E. Gotsman, E. Levin and U. Maor, Nucl. Phys. A696, 851 (2001)
[arXiv:hep-ph/0102321]; Eur. Phys. J. C27, 411 (2003) [arXiv:hep-ph/0209074]; K. Golec-Biernat, L. Motyka and
A.Stasto, Phys. Rev. D65, 074037 (2002) [arXiv:hep-ph/0110325]; E. Iancu, K. Itakura and S. Munier, Phys. Lett.B590 (2004) 199 [arXiv:hep-ph/0310338]. K. Rummukainen and H. Weigert, Nucl. Phys. A739, 183 (2004)
[arXiv:hep-ph/0309306]; K. Golec-Biernat and A. M. Stasto, Nucl. Phys. B668, 345 (2003)
[arXiv:hep-ph/0306279]; E. Gotsman, M. Kozlov, E. Levin, U. Maor and E. Naftali, Nucl. Phys. A742, 55 (2004)
[arXiv:hep-ph/0401021]; K. Kutak and A. M. Stasto, Eur. Phys. J. C41, 343 (2005)
[arXiv:hep-ph/0408117]; G. Chachamis, M. Lublinsky and A. Sabio Vera, Nucl. Phys. A748, 649 (2005)
[arXiv:hep-ph/0408333]; J. L. Albacete, N. Armesto, J. G. Milhano, C. A. Salgado and U. A. Wiedemann, Phys. Rev.
D71, 014003 (2005) [arXiv:hep-ph/0408216]; E. Gotsman, E. Levin, U. Maor and E. Naftali, Nucl. Phys. A750 (2005)
391 [arXiv:hep-ph/0411242].
[9] A. H. Mueller, A. I. Shoshi and S. M. H. Wong, Nucl. Phys. B 715, 440 (2005) [arXiv:hep-ph/0501088].
[10] E. Levin and M. Lublinsky, Nucl. Phys. A 763, 172 (2005) [arXiv:hep-ph/0501173].
[11] E. Iancu and D. N. Triantafyllopoulos, Nucl. Phys. A756, 419 (2005) [arXiv:hep-ph/0411405]; Phys. Lett. B610, 253
(2005) [arXiv:hep-ph/0501193].
[12] A. Kovner and M. Lublinsky, Phys. Rev. D 71, 085004 (2005) [arXiv:hep-ph/0501198].
[13] Y. Hatta, E. Iancu, L. McLerran, A. Stasto and D. N. Triantafyllopoulos, Nucl. Phys. A 764, 423 (2006)
[arXiv:hep-ph/0504182].
[14] E. A. Kuraev, L. N. Lipatov, and F. S. Fadin, Sov. Phys. JETP 45, 199 (1977); Ya. Ya. Balitsky and L. N. Lipatov,
Sov. J. Nucl. Phys. 28, 22 (1978).
[15] J. Bartels, M. Braun and G. P. Vacca, Eur. Phys. J. C40, 419 (2005) [arXiv:hep-ph/0412218] ; J. Bartels and
C. Ewerz, JHEP 9909, 026 (1999) [arXiv:hep-ph/9908454] ; J. Bartels and M. Wusthoff, Z. Phys. C66, 157 (1995) ;
A. H. Mueller and B. Patel, Nucl. Phys. B425, 471 (1994) [arXiv:hep-ph/9403256]; J. Bartels, Z. Phys. C60, 471
(1993).
[16] M. A. Braun, Phys. Lett. B632 (2006) 297 [arXiv:hep-ph/0512057]; arXiv:hep-ph/0504002 ; Eur. Phys. J. C16, 337
(2000) [arXiv:hep-ph/0001268]; Phys. Lett. B 483 (2000) 115 [arXiv:hep-ph/0003004]; Eur. Phys. J. C 33 (2004) 113
[arXiv:hep-ph/0309293]; Eur. Phys. J. C6, 321 (1999) [arXiv:hep-ph/9706373]; M. A. Braun and G. P. Vacca, Eur.
Phys. J. C6, 147 (1999) [arXiv:hep-ph/9711486].
[17] E. Iancu, A. H. Mueller and S. Munier, Phys. Lett. B 606 (2005) 342 [arXiv:hep-ph/0410018]; E. Brunet, B. Derrida,
A. H. Mueller and S. Munier, arXiv:cond-mat/0603160; Phys. Rev. E73 (2006) 056126 [arXiv:cond-mat/0512021].
[18] R. Enberg, K. Golec-Biernat and S. Munier, Phys. Rev. D72 (2005) 074021 [arXiv:hep-ph/0505101]. S. Munier,
arXiv:hep-ph/0608036.
[19] M. Kozlov, E. Levin and A. Prygarin, arXiv:hep-ph/0606260.
[20] M. Ciafaloni, D. Colferai, G.P. Salam and A.M. Stasto, Phys. Rev. D68 (2003) 114003; M. Ciafaloni, D. Colferai,
G. P. Salam and A. M. Stasto, Phys. Lett. B541, 314 (2002) [arXiv:hep-ph/0204287]; Phys. Rev. D66, 054014 (2002)
[arXiv:hep-ph/0204282]; V.S. Fadin and L.N. Lipatov, Phys. Lett. B429 (1998) 127
G. Camici and M. Ciafaloni,Phys. Lett. B430 (1998) 349.
[21] O.V. Kancheli, Sov. JETP Lett. 18, 274 (1973), in Caneschi, L. (ed.): Regge theory of low-p(T) hadronic interactions,p. 312, North-Holland, 1989; Pisma Zh. Eksp. Teor. Fiz. 22 (1975) 491.
[22] A. H. Mueller, Nucl. Phys. B415, 373 (1994); ibid B437, 107 (1995).
[23] A. H. Mueller and B. Patel, Nucl. Phys. B425, 471 (1994); A. H. Mueller and G. P. Salam, Nucl. Phys. B475, 293
(1996), [arXiv:hep-ph/9605302]; G. P. Salam, Nucl. Phys. B461, 512 (1996); E. Iancu and A. H. Mueller, Nucl. Phys.
A730 (2004) 460, 494, [arXiv:hep-ph/0308315],[arXiv:hep-ph/0309276].
33
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
35/36
[24] E. Levin and A. Prygarin, The BFKL pomeron calculus in zero transverse dimension: Summation of the Pomeron
loops and the generating functional for the multiparticle production processes, arXiv:hep-ph/0701178.
[25] E. Levin and K. Tuchin, Nucl. Phys. A 693 (2001) 787 [arXiv:hep-ph/0101275] ; Nucl. Phys. A 691 (2001) 779
[arXiv:hep-ph/0012167], Nucl. Phys. B 573 (2000) 833 [arXiv:hep-ph/9908317] .
[26] E. Levin and M. Lublinsky, Nucl. Phys. A730, 191 (2004) [arXiv:hep-ph/0308279].[27] E. Levin and M. Lublinsky, Phys. Lett. B607, 131 (2005) [arXiv:hep-ph/0411121].
[28] S. Bondarenko, L. Motyka, A. H. Mueller, A. I. Shoshi and B. W. Xiao, On the equivalence of Reggeon field theory in
zero transverse dimensions and reaction-diffusion processes, , arXiv:hep-ph/0609213. P. Rembiesa and A. M. Stasto,
Nucl. Phys. B725 (2005) 251 [arXiv:hep-ph/0503223] ; A. I. Shoshi and B. W. Xiao, Phys. Rev. D 73, 094014 (2006)
[arXiv:hep-ph/0512206]; Phys. Rev. D 75 (2007) 054002 [arXiv:hep-ph/0605282]; M. Kozlov and E. Levin, Nucl. Phys.
A 779 (2006) 142 [ arXiv:hep-ph/0604039]; M. Kozlov, E. Levin, V. Khachatryan and J. Miller, The BFKL pomeron
calculus in zero transverse dimensions: Diffractive processes and survival probability for central diffractive production,
arXiv:hep-ph/0610084.
[29] D. Amati, L. Caneschi and R. Jengo, Nucl. Phys. B 101 (1975) 397 V. Alessandrini, D. Amati and R. Jengo, Nucl.
Phys. B 108 (1976) 425; R. Jengo, Nucl. Phys. B 108 (1976) 447; D. Amati, M. Le Bellac, G. Marchesini and
M.Ciafaloni, Nucl. Phys. B112 (1976) 107; D. Amati, G. Marchesini, M.Ciafaloni and G. Parisi, Nucl. Phys. B114
(1976) 483; M. Ciafaloni, M. Le Bellac and G. C. Rossi, Nucl. Phys. B 130 (1977) 388.
[30] M. Kozlov, E. Levin, V. Khachatryan and J. Miller, arXiv:hep-ph/0610084.
[31] P. Grassberger and K. Sundermeyer, Phys. Lett. B77 (1978) 220.
[32] K. G. Boreskov, Probabilistic model of Reggeon field theory, arXiv:hep-ph/0112325 and reference therein.
[33] V. A. Abramovsky, V. N. Gribov and O. V. Kancheli, Yad. Fiz. 18, 595 (1973) [Sov. J. Nucl. Phys. 18, 308 (1974)].
[34] Y. V. Kovchegov and E. Levin, Nucl. Phys. B 577 (2000) 221 [arXiv:hep-ph/9911523].
[35] Y. V. Kovchegov, Phys. Rev. D 72 (2005) 094009 [arXiv:hep-ph/0508276].
[36] M. Kozlov and E. Levin, Nucl. Phys. A739 (2004) 291 [arXiv:hep-ph/0401118].
[37] E. Levin, Nucl. Phys. A763, 140 (2005), arXiv:hep-ph/0502243.
[38] M. Kozlov and E. Levin, Solution for the BFKL pomeron calculus in zero transverse dimensions,arXiv:hep-ph/0604039.
[39] S. Bondarenko, L. Motyka, A. H. Mueller, A. I. Shoshi and B. W. Xiao, arXiv:hep-ph/0609213.
[40] I. Gradstein and I. Ryzhik, Tables of Series, Products, and Integrals, Verlag MIR, Moskau,1981.
[41] M. Abramowitz and I. Stegun, Handbook of Mathematical Function, Dover Publications, Inc., New York, 1972.
[42] L. N. Lipatov, Phys. Rept. 286, 131 (1997) [arXiv:hep-ph/9610276]; Sov. Phys. JETP 63, 904 (1986) and references
therein.
[43] H. Navelet and R. Peschanski, Nucl. Phys. B634, 291 (2002) [arXiv:hep-ph/0201285]; Phys. Rev. Lett. 82, 137 (1999),
[arXiv:hep-ph/9809474]; Nucl. Phys. B507, 353 (1997) [arXiv:hep-ph/9703238].
[44] Y. Hatta and A. H. Mueller, Correlation of small-x gluons in impact parameter space, arXiv:hep-ph/0702023.
[45] A. H. Mueller and A. I. Shoshi, Nucl. Phys. B692 (2004) 175 [arXiv:hep-ph/0402193].
[46] J. Bartels and E. Levin, Nucl. Phys. B387 (1992) 617.
[47] A. H. Mueller and D. N. Triantafyllopoulos, Nucl. Phys. B640 (2002) 331
[arXiv:hep-ph/0205167]; D. N. Triantafyllopoulos, Nucl. Phys. B 648 (2003) 293 [arXiv:hep-ph/0209121].
[48] S. Munier and R. Peschanski, Phys. Rev. D70 (2004) 077503; D69 (2004) 034008 [arXiv:hep-ph/0310357]; Phys. Rev.
Lett. 91 (2003) 232001 [arXiv:hep-ph/0309177].
34
-
8/3/2019 E. Levin, J. Miller and A. Prygarin-Summing Pomeron loops in the dipole approach
36/36
[49] A. H. Mueller, Nucl. Phys. B558 (1999) 285, [arXiv:hep-ph/9904404].
[50] M. Kozlov, E. Levin and A. Prygarin, The BFKL pomeron calculus in the dipole approach, arXiv:0704.2124
[hep-ph] ; S. Bondarenko, Langevin equation in effective theory of interacting QCD pomerons in the limit of large Nc,
arXiv:hep-ph/0703045.
[51] http://mathworld.wolfram.com/ConfluentHypergeometricLimitFunction.html[52] J. Bartels, M. A. Braun, D. Colferai and G. P. Vacca, Eur. Phys. J. C 20, 323 (2001) [arXiv:hep-ph/0102221].
[53] K. J. Golec-Biernat and M. Wusthoff, Phys. Rev. D 59 (1999) 014017 [arXiv:hep-ph/9807513] , Phys. Rev. D 60 (1999)
114023 [arXiv:hep-ph/9903358] ; E. Gotsman, E. Levin, M. Lublinsky, U. Maor, E. Naftali and K. Tuchin, J. Phys. G
27 (2001) 2297 [arXiv:hep-ph/0010198] ; H. Kowalski and D. Teaney, Phys. Rev. D 68 (2003) 114005
[arXiv:hep-ph/0304189] ; J. Bartels, K. J. Golec-Biernat and H. Kowalski, Phys. Rev. D 66 (2002) 014001
[arXiv:hep-ph/0203258].
[54] E. Levin, Nucl. Phys. B 453 (1995) 303 [arXiv:hep-ph/9412345] ; M. A. Braun, Phys. Lett. B 348 (1995) 190
[arXiv:hep-ph/9408261].
[55] Y. V. Kovchegov and H. Weigert, Nucl. Phys. A 789 (2007) 260 [arXiv:hep-ph/0612071] ; Nucl. Phys. A 784 (2007) 188
[arXiv:hep-ph/0609090] ; E. Gardi, J. Kuokkanen, K. Rummukainen and H. Weigert, Nucl. Phys. A 784 (2007) 282
[arXiv:hep-ph/0609087] ; I. Balitsky, Phys. Rev. D 75 (2007) 014001 [arXiv:hep-ph/0609105].